0

i have 2x 3064PQ and connect both QSFP together and create vpc over this QSP connection (i have a lacp from n3k-1 qsfp 1&2 to n3k-2 qsfp1-2)

and here is vpc config for n3k-1 :

vpc domain 1
peer-switch
role priority 20
system-priority 120
peer-keepalive destination 172.x.x.111 source 172.x.x.110
peer-gateway
auto-recovery



interface port-channel20
speed 40000
description VPC-LINK
switchport mode trunk
switchport trunk allowed vlan 802,1609
spanning-tree port type network
vpc peer-link

and here is n3k-2 vpc config :

vpc domain 1
peer-switch
role priority 30
system-priority 120
peer-keepalive destination 172.x.x.110 source 172.x.x.111
peer-gateway
auto-recovery



interface port-channel20
speed 40000
description VPC-LINK
switchport mode trunk
switchport trunk allowed vlan 802,1609
spanning-tree port type network
vpc peer-link

and on both n3k i have bgp session with same upstream router and i am announcing 85.185.x.x/24 with no prepend and equal cost/path to my upstream router, i want to achieve load balancing and some flows enter my network from n3k-1 and some of them enter from n3k-2 , 

so do i config my switches correctly? does l2/l3 packet working perfectly in this senario ?

and at last my both nexus have 2x 10g LACP with my access switches (which my Top of rack switches are connected to them) and my vlans are permiited on both uplinks,

Thank you.

1
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer.
    – Ron Maupin
    Dec 16 '20 at 23:32
2

I'd suggest adding ip arp synchronize under the VPC configuration. It would also be smart to add an auto-recovery delay (both standard and reload) to account for letting a returning switch to stabilize its routing table before reasserting itself in the pair.

There's a bunch of other configuration that's also important - the HSRP / VRRP setup (priorities need to match VPC role), spanning tree values (...priorities need to match).

You also might want to consider either directly peering the switches (over a separate physical interface) or setting up object tracking to potentially pull the downstream links out of service if the upstream drops. Keep in mind that without such measures that you're going to black-hole a bunch of traffic if you lose upstream peering on one of the switches.

The other thing we can't say based on the info provided is the overall topology. Ideally nothing should be singly connected to a VPC pair - and if something is, you need to account for orphan port handling.

3
  • thank you for your clean explaination! so for confirmation if i use auto-recoery and IP arp sync then i can load balance my traffic, in that scenario i have explained?
    – Blackmetal
    Oct 6 '20 at 7:41
  • 1
    For downstream : Within the limits of whatever upstream is multi-pathing back to you. If your upstream sends all of your inbound traffic to one side, then that’s where forwarding to your hosts will occur. For upstream traffic, the edge devices will send traffic to an HSRP address that’s on the other side of a port-channel. Load balancing will thus be a function of how the edge switch accomplishes multi-pathing over said channel. If it’s purely MAC based then it might be pretty unbalanced. If it uses L3/L4 data for hashing it could be pretty good.
    – rnxrx
    Oct 6 '20 at 19:09
  • i just have 2x 10g lacp in first n3k and 2x 10g lacp in second n3k and in upstream switch side there no other special config such as hsrp ,... so in this way i think some flows should enter first n3k and some flows will enter n3k and it will be approximately but if i have 10g in single flow then it maybe on first n3k or it maybe on second n3k , am i right?
    – Blackmetal
    Oct 6 '20 at 19:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.